I. INTRODUCTION

Top Previous Next References

Mitogen-activated protein kinase (MAPK) signal transduction pathways are among the most widespread mechanisms of eukaryotic cell regulation. All eukaryotic cells possess multiple MAPK pathways, each of which is preferentially recruited by distinct sets of stimuli, thereby allowing the cell to respond coordinately to multiple divergent inputs. Mammalian MAPK pathways can be activated by a wide variety of different stimuli acting through diverse receptor families, including hormones and growth factors that act through receptor tyrosine kinases [e.g., insulin, epidermal growth factor (EGF), platelet-derived growth factor (PDGF), and fibroblast growth factor (FGF)] or cytokine receptors (e.g., growth hormone) to vasoactive peptides acting through G protein-coupled, seven-transmembrane receptors (e.g., ANG II, endothelin), transforming growth factor (TGF)--related polypeptides, acting through Ser-Thr kinase receptors, inflammatory cytokines of the tumor necrosis factor (TNF) family and environmental stresses such as osmotic shock, ionizing radiation and ischemic injury. MAPK pathways, in turn, coordinate activation of gene transcription, protein synthesis, cell cycle machinery, cell death, and differentiation. Accordingly, these pathways exert a profound effect on cell physiology (120, 165, 195).

All MAPK pathways include central three-tiered "core signaling modules" (Fig. 1) in which MAPKs are activated by concomitant Tyr and Thr phosphorylation within a conserved Thr-X-Tyr motif in the activation loop of the kinase domain subdomain VIII. MAPK phosphorylation and activation are catalyzed by a family of dual specificity kinases referred to as MAPK/extracellular signal-regulated kinase (ERK)-kinases (MEKs or MKKs). MEKs, in turn, are regulated by Ser/Thr phosphorylation, also within a conserved motif in kinase domain subdomain VIII, catalyzed by any of several protein kinase families collectively referred to as MAPK-kinase-kinases (MAP3Ks). MAPK core signaling modules are themselves regulated by a wide variety of upstream activators and inhibitors (120, 165, 195).

View larger version (27K): [in this window] [in a new window]   Fig. 1. The mitogen-activated protein kinase (MAPK) core signaling module. Divergent inputs feed into core MAPK-kinase-kinase (MAP3K)  MAPK-kinase (MEK)  MAPK core pathways which then recruit appropriate responses.

The notion of multiple parallel MAPK signaling cascades was first appreciated from studies of simple eukaryotes such as the budding yeast Saccharomyces cerevisiae. To date, six S. cerevisiae MAPK signaling pathways have been identified. These have been reviewed elsewhere (120). The features of yeast MAPK pathways as well as early biochemical studies have revealed several common principles shared by all MAPK pathways.

1) MAPKs are proline-directed kinases; however, substrate selectivity is often conferred by specific MAPK docking sites present on physiological substrates. The proline-directed substrate specificity of the MAP kinases was established using peptide substrates corresponding to the sequences surrounding Thr 669 of the EGF receptor (Glu-Leu-Val-Glu-Pro-Leu-Thr-Pro-Ser-Gly-Glu-Ala-Pro- Asn-Gln-Ala-Leu-Leu-Arg) and the site on myelin basic protein (Ala-Pro-Arg-Thr-Pro-Gly-Gly-Arg) phosphorylated in vitro by the 42-kDa insulin- and mitogen-stimulated MAPK ERK-2. Systematic variation in the sequences surrounding the single Thr phosphorylation site in the EGFR peptide and the single Thr in the MBP peptide, verified the essential role of the proline at immediately COOH terminal to the phosphoacceptor site (50, 55, 84). However, this proline directedness is not sufficient to account for the high degree of substrate selectivity manifested by different MAPK subgroups. The Michaelis constant (K_m) for native protein substrates is usually several orders of magnitude lower than the K_m for synthetic peptides corresponding in amino acid sequence to the region immediately surrounding the phosphorylation site on these substrates. It has become apparent that most, if not all, of the physiological substrates of the MAPKs possess specific binding sites, often at considerable distance from the phosphorylation site in the primary sequence, that allow for a strong interaction with select MAPK subfamilies to the exclusion of others (64, 144, 301, 351). In turn, MAPKs themselves possess complementary docking sites that interact with the MAPK binding domains on substrate proteins (143, 301). This confers a striking substrate specificity on a family of protein kinases with an otherwise apparently broad in vitro substrate profile, certainly as portrayed by synthetic peptides.

2) There are signaling components with more than one biological function/signaling components under multiple forms of regulation. Within MAPK core signaling modules, there are instances wherein individual elements can function promiscuously in several pathways. Conversely, MAPK pathway components are often subject to regulation by multiple inputs. For example, the S. cerevisiae MAP3K Ste11p functions as part of the mating pheromone response pathway and the osmosensing pathway (20, 242), while the mammalian MAP3K tumor progression locus-2 (Tpl-2) can activate the mitogenic ERK and three stress-activated MAPK pathways (264). Conversely, the yeast osmosensing MEK Pbs2p can be regulated by three MAP3Ks: Ssk2p, Ssk22p, and Ste11p while the mammalian stress-regulated MEK stress-activated protein kinase/ERK-kinase-1 (SEK1)/MAPK-kinase-4 (MKK4) is a putative substrate for at least 10 known MAP3Ks.

3) Pathway organization is mediated by scaffolding proteins. Given the extraordinary complexity and diversity of MAPK regulation and function, it is critical that the efficiency and selectivity of MAPK pathways be preserved. Scaffold proteins bind and sequester select MAPK pathway components, and thereby help to maintain pathway integrity and to permit the coordinated and efficient activation and function of MAPK components in response to specific types of stimuli (234). Some yeast signaling pathways include distinct scaffolding proteins that bind and segregate groups of signaling components. Alternatively, in some MAPK pathways, the signaling components themselves possess intrinsic scaffolding properties (see Fig. 7). Thus Ste5p of the yeast mating pheromone pathway is a scaffolding protein that selectively binds a MAP3K (Ste11p), a MEK (Ste7p), and a MAPK (Fus3p) and couples them to upstream activators. Thus, although Ste11p can function in both the mating and osmosensing pathways, it selectively activates different MEKs in each pathway: Ste7p for the mating pathway and Pbs2p for the osmosensing pathway, due in part to the fact that Ste5p maintains signaling pathway specificity by binding selectively Ste7p and not Pbs2p (120, 242).

Conversely, Pbs2p, in addition to serving as a MEK, has intrinsic scaffold properties, selectively binding Ste11p and the osmosensing MAPK Hog1p. Pbs2p does not bind Fus3p, and thus Pbs2p maintains signaling pathway integrity by interacting specifically with Hog1p and not with Fus3p or Kss1p (3, 11). Likewise, the mammalian scaffold proteins c-Jun NH₂-terminal kinase (JNK) interacting proteins (JIPs)-1 and -2 and JNK/stress-activated protein kinase (SAPK)-associated protein-1 (JSAP1/JIP3), like Ste5p, couple elements of different SAPK core signaling modules, while mammalian stress-activated protein kinase (SAPK)/ERK-kinase-1 (SEK1), like Pbs2p, possess both the properties of a protein kinase and a scaffold protein (72, 136, 155, 336, 342, 355).

4) MAP3K regulation is by membrane recruitment, oligomerization, and phosphorylation. MAP3K regulation represents the entry point to MAPK pathways and is accordingly complex. In addition to scaffold proteins mentioned above, actual activation of MAP3Ks has been shown to involve three key phenomena: recruitment to the membrane, typically mediated by an upstream activating protein; homoligomerization, often within a multiprotein complex containing additional regulatory components; and phosphorylation by upstream kinases. The mitogenic MAP3K Raf is illustrative of all three properties. Thus it is recruited to the membrane by Ras (195), where it undergoes regulatory phosphorylation catalyzed in part by p21-activated kinase-3 (160) and oligomerization (190). All three events are necessary for activation (see sect. IIIA1). Stress-regulated MAP3Ks are likely similarly regulated.

	II. THE STRESS-ACTIVATED PROTEIN KINASE/c-JUN NH2-TERMINAL KINASE, p38 AND ERK5/BIG MITOGEN-ACTIVATED PROTEIN KINASE-1 PATHWAYS: THE THREE MAJOR MAMMALIAN MITOGEN-ACTIVATED PROTEIN KINASE SIGNALING PATHWAYS ACTIVATED BY ENVIRONMENTAL STRESS AND INFLAMMATORY CYTOKINES: CORE PATHWAYS AND THEIR TARGETS

Top Previous Next References

The insulin/mitogen-regulated extracellular signal-regulated kinase (ERK) pathway was the first mammalian MAPK pathway to be identified. This pathway is largely regulated by the monomeric GTPase Ras which recruits MAP3Ks of the Raf family to activate two MEKs: MEK1 and MEK2. These, in turn, activate the ERKs. The biochemistry, biology, and regulation of the ERKs have been reviewed exhaustively elsewhere (10, 11, 195).

In recent years, it has become clear that, as with yeast, multiple parallel mammalian MAPK pathways exist and that most of these, in conjunction with the nuclear factor-B (NF-B) pathway, are pivotal to stress and inflammatory responses rather than to mitogen responses. As mammalian stress-activated signaling pathways are elucidated, it is becoming evident that these pathways will be important targets for novel anti-inflammatory drugs.

Indications that mammalian cells possessed several MAPK pathways came shortly after the identification of the ERKs. The protein synthesis inhibitor cycloheximide, when administrated to rats, can elicit the in vivo activation of ribosomal S6 phosphorylation, and, in fact, this strategy was used to activate p70 S6 kinase in vivo before purification (10). It was originally suggested that p70 S6 kinase, like MAPK-activated protein kinase-1 (MAPKAP-K1)/ribosomal S6 kinase (Rsk), a known ERK substrate with which p70 S6 kinases shares some superficial enzymological similarities (85, 254, 292), would be activated by ERKs. However, the ERKs were unable to activate p70 S6 kinase in vitro, suggesting that cycloheximide activated novel kinases including proline-directed protein kinases (164). This idea was supported by the demonstration that injection of cycloheximide into rats activated a novel Ser/Thr kinase activity that could phosphorylate microtubule-associated protein-2 (MAP2). Purification of this kinase revealed a 54-kDa polypeptide, initially named p54-MAP2 kinase, or p54 (164, 165). p54 could be inactivated with Tyr or Ser/Thr phosphatases, indicating that, like the ERKs, the p54 kinase required concomitant Tyr and Ser/Thr phosphorylation for activity (167, 165); moreover, studies with synthetic peptide substrates showed that p54 was proline directed, but phosphorylated Ser/Thr-Pro motifs with a specificity distinct for that of the p42/p44 MAPKs (220).

The specificity of p54 toward native protein substrates clearly differed from that of the ERKs. In particular, p54 was unable to activate MAPKAP-K1/Rsk in vitro under conditions wherein ERK-mediated activation of MAPKAP-K1/Rsk was observed (164). Of note, p54 was also unable to activate p70 S6 kinase in vitro, although both p54 and the ERKs could phosphorylate p70 S6 kinase in an autoinhibitory segment. p70 S6 kinase activation was subsequently shown to require additional phosphatidylinositol (PI) 3-kinase-dependent steps (10). Most importantly, p54 was able to phosphorylate the c-Jun transcription factor at two sites (Ser-63 and Ser-73) implicated in regulation of c-Jun and AP-1 trans-activation function (244), at a substantially higher rate than was observed for the p42/p44 MAPKs.

The SAPKs were cloned independently by two groups: Kyriakis et al. (166) used the amino acid sequence of tryptic peptides derived from purified p54 to design specific PCR primers, whereas Dérijard et al. (68) used a pure PCR strategy employing degenerate primers derived from regions conserved among known MAPKs. The generation of specific antibodies enabled an analysis of the regulation of endogenous p54 by extracellular stimuli. Assay of p54 kinase activity immunoprecipitated from extracts of cells subjected to various treatments revealed that, in contrast to the p42/p44 MAPKs, p54 was not strongly activated in most cells by mitogens such as insulin, EGF, PDGF, or FGF. In contrast, p54 was vigorously activated by a variety of noxious treatments such as heat shock, ionizing radiation, oxidant stress, DNA damaging chemicals (topoisomerase inhibitors and alkylating agents), reperfusion injury, mechanical shear stress, and, of course, protein synthesis inhibitors (cycloheximide and anisomycin) (68, 165, 166, 237).

In rationalizing the significance of p54 activation by environmental stresses, the potent activation of p54 by tunicamycin was conceptually important; tunicamycin inhibits N-linked protein glycosylation and leads to the accumulation of misfolded proteins exclusively within the lumen of the endoplasmic reticulum (ER). The ability of an ER-localized perturbation to activate the cytosolic p54 kinase was one of the first clear-cut examples of stress-activated signal transduction through a protein kinase (166). This formulation suggested that stress-activated cytokines were likely regulators of p54 activity, a surmise rapidly borne out by the demonstration that p54 is strongly activated by all the inflammatory cytokines of the TNF family [TNF, interleukin (IL)-1, CD40 ligand, CD27 ligand, Fas ligand, receptor activator of NF-B, RANK ligand, etc.] as well as by vasoactive peptides (endothelin and ANG II) (4, 16, 19, 68, 166, 178, 275, 369). p54 has since been renamed, and the nomenclature of this family of kinases is somewhat confusing (Table 1). Two systems are generally accepted: stress-activated protein kinase (SAPK) in reference to the regulation of these kinases by environmental stress and inflammation and c-Jun NH₂-terminal kinase (JNK) in reference to the phosphorylation by these kinases of the c-Jun NH₂-terminal trans-activation domain.

The SAPKs are encoded by at least three genes: SAPK/JNK2, SAPK/JNK3, and SAPK/JNK1 (111, 143, 166) (Table 1). As with all MAPKs, each SAPK isoform contains a characteristic Thr-X-Tyr phosphoacceptor loop in subdomain VIII of the protein kinase catalytic domain. Whereas the ERK sequence is Thr-185-Glu-Tyr-187 (ERK2) or Thr-203-Glu-Tyr-205 (ERK1), that of the SAPKs is Thr-183-Pro-Tyr-185. The expression of each SAPK gene is further diversified by differential hnRNA splicing. Splicing within the catalytic domain at a region spanning subdomains IX and X results in type 1 and type 2 SAPKs ( and  JNKs, respectively, Table 1), whereas splicing at the extreme COOH terminus yields 54-kDa (p54) and 46-kDa (p46) polypeptides (type 2 and type 1 JNKs, respectively, Table 1); thus at least 12 polypeptides have been identified. The significance of the COOH-terminal isoforms is not clear, but the type 1 and type 2 kinases differ in their substrate binding affinities (63, 111, 143, 166).

The p38 MAPKs are a second mammalian stress-activated MAPK family. Originally described as a 38-kDa polypeptide that underwent Tyr phosphorylation in response to endotoxin treatment and osmotic shock (117), p38 (the -isoform) was purified by anti-phosphotyrosine immunoaffinity chromatography; cDNA cloning revealed that p38 was the mammalian MAPK homolog most closely related to HOG1, the osmosensing MAPK of S. cerevisiae. Most notably, the p38s, like Hog1p, contain the phosphoacceptor sequence Thr-Gly-Tyr (117, 120). Independently and contemporaneously, two groups identified p38 as a kinase activated by stress and IL-1 that could phosphorylate and activate MAPK-activated protein kinase-activated protein kinase-1 (MAPKAP kinase-2, see sect. IIE1), a novel Ser/Thr kinase implicated in the phosphorylation and activation of the small 27-kDa heat shock protein HSP27 (95, 259).

Of potential clinical importance, p38 was also purified and cloned as the polypeptide receptor for a class of experimental pyridinyl-imidazole anti-inflammatory drugs, the cytokine-suppressive anti-inflammatory drugs (CSAIDs), the most extensively characterized of which is the compound SB203580 (174). CSAIDs were originally identified in a screen for compounds that could inhibit the transcriptional induction of TNF and IL-1 during endotoxin shock (174). The basis for the efficacy of these compounds as anti-inflammatory agents was their ability to bind and directly inhibit a subset of the p38s, thereby blocking p38-mediated activation of AP-1, a trans-acting factor required for TNF and IL-1 induction (174). Finally, a shorter hnRNA splicing isoform of p38 was isolated as a kinase that could interact with the Myc binding partner Max (362) (see sect. IIE2). With the identification of additional p38 isoforms, four p38 genes are now known (Table 1): the original isoform, here referred to as p38 [also called CSAIDs binding protein (CSBP) and, somewhat confusingly, SAPK2a], p38 (also called SAPK2b, and p38-2), p38 (also called SAPK3 and ERK6), and p38 (also called SAPK4) (95, 106, 117, 139, 140, 173, 174, 207, 259, 288, 327, 363).

In vitro assays demonstrated that only p38 and p38 are inhibited by CSAIDs; p38 and p38 are completely unaffected by these drugs in vitro or in transfected cells (106). The basis for this inhibition was revealed in the crystal structure of p38 complexed with the SB203580. Thr-106 in the hinge of the p38 ATP binding pocket interacts with a fluorine atom in the SB203580 structure. This orients the drug to interact with His-107 and Leu-108 of the ATP binding pocket (86, 110). Substitution of Thr-106, alone or in combination with His-107 or Leu-108, with the corresponding, more bulky residues from p38 or p38 (Met, and Pro or Phe, respectively, in both cases) abolishes SB203580 binding. Conversely, if the amino acid of p38, p38, or even SAPK which corresponds to p38 Thr106 is replaced with Thr, the resulting mutants display at least partial sensitivity to SB203580 (86, 110).

Like the SAPKs, the p38s are strongly activated in vivo by environmental stresses and inflammatory cytokines and are inconsistently activated by insulin and growth factors. In almost all instances, the same stimuli that recruit the SAPKs also recruit the p38s (165). One exception is ischemia-reperfusion. SAPKs are not activated during ischemia, but rather during reperfusion, whereas the p38s are activated during ischemia and remain active during reperfusion (22, 165, 237). The basis for this difference is unknown.

The novel MEK MEK5 was cloned by degenerate PCR as part of an effort to identify new MAPK pathways and regulators (83, 367). ERK5, a MEK5 substrate, was cloned as part of a two-hybrid screen that employed MEK5 as bait (367). ERK5 is a ~90-kDa MAPK of which only one mammalian homolog is known. ERK5 has the sequence Thr-Glu-Tyr in its phosphoacceptor loop. The NH₂-terminal kinase domain of ERK5 is followed by an extensive COOH-terminal tail of unknown function that contains several consensus proline-rich motifs indicative of binding sites for proteins with SH3 domains (367).

The stimuli that recruit ERK5 have not been comprehensively characterized; however, ERK5 can be substantially activated by environmental stresses such as oxidant stress (peroxide) and osmotic shock (sorbitol), but not by vasoactive peptides or inflammatory cytokines (TNF) (1). ERK5 may also lie downstream of receptor Tyr kinases. Although mitogen activation of ERK5 was not initially observed (1), EGF activation of ERK5 has been subsequently documented (38).

A) MITOGEN-ACTIVATED PROTEIN KINASE-ACTIVATED PROTEIN KINASES (MAPKAP-KS)-2 AND -3. MAPKAP kinase-2 (MAPKAP-K2) and the structurally related MAPKAP-K3 (also called three pathway regulated kinase or 3PK) are a small family of Ser/Thr kinases that consist of an NH₂-terminal regulatory domain and a COOH-terminal kinase domain (132, 169, 203, 283, 289, 290). They are unrelated to the MAPKAP-K1s/Rsks, targets of the ERKs (11, 85). Along with p38-regulated and activated kinase (PRAK, see below), MAPKAP-K2 and MAPKAP-K3 phosphorylate the small heat shock protein HSP27 (132, 169, 203, 283, 289, 290). Nonphosphorylated HSP27 normally exists in high-molecular-weight multimers that serve as molecular chaperones. Phosphorylation of HSP27 by MAPKAP kinase-2/3 at Ser-15, Ser-78, Ser-82, and Ser-90 coincides with the dissociation of HSP27 into monomers and dimers, and with the redistribution of HSP27 to the actin cytoskeleton (132). In peroxide-treated human umbilical vein endothelial cells, this redistribution of HSP27 may contribute to triggering the reorganization of F-actin into stress fibers, thereby affecting cell motility (132, 169). MAPKAP-K2/3-catalyzed phosphorylation of HSP27 at Ser-90 appears necessary for this process, and mutation of Ser-90 to Ala prevents stimulus-induced changes in HSP27 oligomerization (169).

Early studies had indicated that purified ERK1 could phosphorylate and activate MAPKAP-K2 in vitro (289). However, ERKs likely do not represent the physiological MAPKAP-K2 kinases. MAPKAP-K2 is activated not by insulin or mitogens, conditions wherein ERKs are strongly activated, but by stresses and inflammatory cytokines, conditions wherein ERKs are not appreciably activated (95, 259). MAPKAP-K2 is phosphorylated and activated by p38 and p38 (but not by p38 or p38) (60, 95, 259) (Fig. 2). Activation of MAPKAP-K2 is a multistep process wherein phosphorylation of Thr-25, catalyzed by either p38 or p38, gates subsequent phosphorylation of Thr-222 and Ser-272 in the kinase activation loop, a reaction also catalyzed by either p38 or p38. Together, these phosphorylations, accompanied by an additional autophosphorylation at Thr-334, result in activation of MAPKAP-K2 (15). Consistent with regulation by p38 and p38, MAPKAP-K2 activation and HSP27 phosphorylation are inhibited by CSAIDs (15, 132, 169, 203).

View larger version (28K): [in this window] [in a new window]   Fig. 2. Downstream protein kinases and their targets regulated by p38s. Note that MSKs and MNKs are also regulated by the extracellular signal-regulated kinases (ERKs).

MAPKAP-K3 can also phosphorylate HSP27 (169). Although it has been shown that MAPKAP-K3 can be activated in vivo, in overexpression experiments, by the ERK, SAPK, and p38 pathways (189), endogenous MAPKAP-K3 is only significantly activated by stresses and inflammatory cytokines, and in a manner that can be completely inhibited with CSAIDs, suggesting that MAPKAP-K3 is, in fact, stress and not mitogen activated, and that p38 and/or p38 are the major MAPKAP-K3 kinases in vivo (169).

B) P38-REGULATED/ACTIVATED KINASE. p38-regulated/activated kinase (PRAK) is a ~50-kDa Ser/Thr kinase with a similar overall structure to MAPKAP-K2, -K3, and the MAPK-interacting kinases (MNKs, see below). Accordingly, PRAK consists of an NH₂-terminal regulatory domain and a COOH-terminal kinase domain. Like MAPKAP -K2 and -K3, PRAK is activated selectively in response to stress and inflammatory cytokines and is not detectably activated by mitogens. PRAK can be activated in vivo and in vitro by p38 and p38, and consistent with this, PRAK activation can be blocked with CSAIDs (224). Phosphorylation of PRAK catalyzed by p38s is at Thr-182 in the kinase domain activation loop (224). Once activated, PRAK can phosphorylate HSP27 at the physiologically relevant sites, and in-gel kinase assays indicate that PRAK is an important stress-activated HSP27 kinase (224) (Fig. 2).

C) MNKS. Cellular mRNAs contain a 5'-cap structure, the N⁷-methylguanosine cap. The N⁷-methylguanosine-binding protein eIF-4E recruits mRNAs onto a scaffold protein eIF-4G, which also binds the RNA helicase eIF-4A. The latter, in collaboration with the RNA binding protein eIF-4B, unwind secondary structure in the mRNA 5'-untranslated segment, thereby facilitating the scanning of the mRNA by the 40S ribosomal complex to the ATG translational initiation site. The complex of eIF-4A, B, G, and E is known as eIF-4F (285).

The eIF-4E-eIF-4G interaction is negatively regulated by the translational repressor protein 4E-binding protein-1 (4E-BP1 also called phosphorylated heat- and acid-stable protein regulated by insulin, PHAS-I). 4E-BP1 is multiply phosphorylated in response to insulin and mitogens resulting in the dissociation of 4E-BP1 from eIF-4E, and the release of eIF-4E for incorporation into the eIF-4F complex. In vivo, the phosphorylation of eIF-4E is strongly inhibited by rapamycin and wortmannin, consistent with the fact that dissociation of 4E-BP1 is regulated by the mammalian target of rapamycin (mTOR) (11, 285) as well as by protein kinases downstream of PI 3-kinase (285).

In addition, eIF-4E itself also undergoes a regulatory phosphorylation at Ser-209 in response to both insulin/mitogen and environmental stress. This phosphorylation increases the affinity of eIF-4E for the 5'-cap by about threefold; crystallographic data also indicate that phosphorylation of eIF-4E favors the binding of 4E to the 5'-cap (285).

MNK-1 and -2 are two closely related kinases that are probably the physiologically relevant eIF-4E Ser-209 kinases. As the name implies, MNKs associate in vivo with MAPKs and are in vitro and in vivo MAPK substrates. MNKs are phosphorylated and activated both by ERKs 1/2 (in response to insulin and mitogens) and by the p38s (in response to stress) (98, 331, 332). The regulation of the MNKs by both the ERKs and p38s indicates that, as with AP-1 (see below), MNKs are a site of integration of stress and mitogenic signaling pathways (Fig. 2).

D) MSK1/2. Mitogen- and stress-activated protein kinases (MSKs)-1 and -2 are a recently identified family of Ser/Thr protein kinases with an overall structure similar to that of the MAPKAP-K1s/Rsks (85), notably, the MSKs possess two tandem protein kinase domains. MSKs1/2 were identified in a search of databases for DNA sequences homologous to p70 S6 kinase. In vitro MSK1 will phosphorylate the synthetic peptide Crosstide (Gly-Arg-Pro-Arg-Thr-Ser-Ser-Phe-Ala-Glu-Gly); however, MSK1 purified from unstimulated cells possesses low Crosstide kinase activity. In contrast, the Crosstide kinase activity of MSK1 is activated >100-fold upon incubation in vitro with ERK2 (Fig. 2), and MSK1 is activated in vivo by mitogens in a manner inhibitable by the MAPK pathway inhibitor PD98059 (66). MSK1 is also a substrate for the p38 MAPKs. Consistent with this, MSK1 is also activated in vivo by environmental stresses [arsenite, ultraviolet (UV) radiation, peroxide], in a manner inhibited by the CSAID SB203580 (66). MSK1 is also activated by TNF; however, the pharmacological inhibition of TNF activation of MSK1 is more complex than the inhibition of mitogen or stress activation of MSK1. In HeLa cells, TNF activates the p38s (and the SAPKs) as well as the ERKs. However, p38 activation is more rapid, reaching a maximum at 5 min, while ERK activation is comparatively slower, reaching an apparent maximum at 15 min. Accordingly, SB203580 completely blocks early (5 min) TNF activation of MSK1 in HeLa cells. After 15 min of TNF stimulation, however, SB203580 inhibition is only partial, and MSK1 can now also be partially inhibited with the ERK pathway inhibitor PD98059 (66) (Fig. 2).

cAMP response element binding protein (CREB) is a bZIP transcription factor that binds and trans-activates genes containing the cAMP response element (CRE-consensus sequence: TGACGTCA). CREB trans-activating activity can be activated by a number of different stimuli including mitogens, stresses, and, of course, agonists that elevate cAMP. Activation of CREB trans-activating activity correlates with phosphorylation at Ser-133. Protein kinase A (PKA) likely represents the major CREB kinase recruited by cAMP (113). In addition, several protein kinases including MAPKAP-K1s/Rsks (343) and MAPKAP -K2 (300) have been shown to possess mitogen- or stress-activated CREB Ser-133 kinase activity. MSK1 is also a potent CREB kinase in vitro, and a substantial amount of evidence indicates that, in fact, MSK1, and not MAPKAP-K1/Rsk or MAPKAP-K2/3, is the physiologically relevant stress- and mitogen-activated CREB kinase (66).

First, the K_cat for MSK1-catalyzed phosphorylation of CREB is nearly 100-fold greater than that for MAPKAP-K1/Rsk-catalyzed CREB phosphorylation. Moreover, the MSK1 polypeptide has a bipartite nuclear localization signal and is localized exclusively in the nucleus, where CREB resides. MAPKAP-K1s/Rsks are largely cytosolic, although they can translocate to the nucleus in response to extracellular stimuli (40). Finally, the pattern of pharmacological inhibition of stress- and mitogen-activated CREB phosphorylation in vivo tightly corresponds to the pattern of pharmacological inhibition of MSK1 activation. Thus activation of CREB by stress and mitogens is blocked by the broad specificity kinase inhibitor Ro318220. MSK1 is strongly inhibited in vitro by Ro318220, whereas MAPKAP-K2 and MAPKAP-K3 are not. Finally, CREB activation in vivo by TNF can be blocked with SB203580 and the ERK inhibitor PD98059 with kinetics that parallel those described above for inhibition of MSK1 (66).

A recent study suggests that MSKs may be involved in the phosphorylation of components involved in chromatin remodeling. Both mitogens and stresses, as they stimulate gene expression, must prompt the loosening of chromatin structure from around target genes. This enables the transcriptional machinery to gain access to genes that are to be expressed (35, 267). This alteration in chromatin structure involves both the acetylation and phosphorylation of histones. Many transcription factors that are phosphorylated in a stimulus-dependent manner can recruit transcriptional coactivators that are histone acetyl transferases. In addition, protein kinase cascades phosphorylate histones themselves (35, 267). The nucleosomal proteins histone H3 and high mobility group-14 (HMG-14) are prime targets of phosphorylation (35, 267, 308). Agonist-stimulated phosphorylation of histone H3 occurs at Ser-10, while HMG-14 is phosphorylated at Ser-6. Recent genetic evidence indicates that mitogenic histone H3 phosphorylation is mediated at least in part by MAPKAP-K1b/Rsk2 inasmuch as cells derived from patients with Coffin-Lowry syndrome, a loss of function mutation in the mapkap-k1b/rsk2 gene, display a substantial deficit in mitogen-stimulated histone H3 phosphorylation (267).

However, it is unclear if MAPKAP-K1b/Rsk2 is the only histone H3 kinase in vivo. Thus both mitogens and stresses can stimulate H3 and HMG-14 phosphorylation in vivo (308), and MAPKAP-K1s/Rsks are not strongly activated by stresses (10, 164). Moreover, MAPKAP-K1s/Rsks are poor HMG-14 kinases in vitro and are not inhibited by the inhibitor H89 which blocks HMG-14 kinase activity in vivo and in vitro. In contrast, MSK1 is strongly inhibited by H89 in parallel with in vivo inhibition of histone H3 and HMG-14 phosphorylation. In addition, in vitro, MSK1 can phosphorylate histone H3 at Ser-10, at a rate comparable to or better than that phosphorylated by MAPKAP-K1s/Rsks in vitro (308).

A) CREB HOMOLOGOUS PROTEIN/GROWTH ARREST AND DNA DAMAGE-155. CREB homologous protein (CHOP)/growth arrest and DNA damage-155 (GADD153) is a bZIP transcription factor of the CREB family (113). CHOP/GADD153 is transcriptionally induced in response to genotoxic and inflammatory stresses. These treatments can also activate the transcriptional regulatory functions of CHOP/GADD153 through stimulus-induced phosphorylation of Ser-78 and Ser-81. CHOP/GADD153 is a transcriptional repressor of certain cAMP-regulated genes and a transcriptional activator of some stress-induced genes. Activation by genotoxins of CHOP/GADD153 mediates in part cell cycle arrest at G₁/S, an important consequence of DNA damage inasmuch as it allows for DNA repair before DNA replication, key to preserving genomic integrity. p38 is a likely stress-activated regulator of CHOP/GADD153 function, given that p38, and not SAPK or ERK, can phosphorylate CHOP/GADD153 at Ser-78 and Ser-81 in vivo and in vitro (325) (Fig. 3).

View larger version (19K): [in this window] [in a new window]   Fig. 3. Regulation of transcription factors by MAPKs. Note that AP-1 regulation involves both the direct phosphorylation of AP-1 components as well as the transcriptional induction of AP-1 components mediated by distinct transcription factor targets of MAPKs.

B) NUCLEAR FACTOR OF ACTIVATED T CELLS. Transcription factors of the nuclear factor of activated T cells (NFAT) family are distantly related to Rel/NF-B (250). In resting cells, NFATs are retained in the cytosol as a consequence of phosphorylation [catalyzed by casein kinase-I and, possibly, glycogen synthase kinase (GSK)-3] at five or six sites (within NFAT4, this comprises a region spanning amino acids 204-215). NFAT phosphorylation affects conformation so as to mask the nuclear localization signal. Agonist-induced Ca²⁺ entry recruits the Ca²⁺-dependent phosphatase calcineurin (phosphatase 2B), which dephosphorylates NFATs, thereby exposing NFAT's nuclear localization signal and triggering NFAT nuclear translocation. Dephosphorylation of NFATs also enhances DNA binding affinity (250). NFATs bind and trans-activate genes with an NFAT cis-acting element (consensus sequence: ^T/_AGGAAAAT). NFAT sites are often located close to AP-1 sites in many promoters, allowing for the cooperative binding and synergistic trans-activation of numerous genes (IL-2, IL-4, IL-5, and CD40L are examples) (250). Calcineurin is a major target of the immunosuppressants FK506 and cyclosporin A, and accordingly, inhibition of NFAT activity is an important consequence of FK506 and cyclosporin A action (250, 270).

Serum factors can substantially inhibit Ca²⁺-mediated nuclear translocation of NFATs. Insofar as casein kinases and GSK3 are not serum stimulated, the serum-dependent inhibition of Ca²⁺-mediated NFAT translocation suggested that serum-responsive kinase cascades might contribute to NFAT inhibition (47, 250). Davis and colleagues (47) showed that the SAPKs could phosphorylate the NFAT family member NFAT4 (also called NFATc3) at Ser-163 and Ser-165. This phosphorylation correlates with an inhibition of stimulus-induced NFAT4 nuclear translocation, and on the basis of this finding, it has been proposed therefore that the SAPK pathway antagonizes NFAT4 action (47) (Fig. 3). However, the significance of these results is somewhat unclear. Zhu et al. (368) showed that NFAT4 mutants in which the putative SAPK phosphoacceptor sites (Ser-163 and Ser-165) have been changed to Ala still show serum-stimulated inhibition of nuclear translocation in many instances. These investigators also observed that the SAPK-specific MAP3K MEK-kinase-1 (see sect. IIG2) could indirectly, and independently of SAPK, block NFAT4 dephosphorylation and activation. MEKK1 inhibition of NFAT translocation occurs whether or not the SAPK phosphorylation sites have been mutated to alanine, and the effect of MEKK1 is not reversed by dominant inhibitors of the SAPK pathway. Apparently, MEKK1 fosters inhibition of NFAT nuclear translocation by stabilizing the association between NFAT4 and the inhibitory NFAT kinase casein kinase-I (368).

Davis and colleagues (48) have also shown that SAPKs can phosphorylate NFATc1 (also called NFAT2). In this instance, phosphorylation is stimulated by phorbol 12-myristate 13-acetate (PMA) and ionomycin and occurs at Ser-117 and -172. Phosphorylation inhibits or delays the accumulation of NFATc1 in the nucleus by blocking the binding of calcineurin, an essential step in NFATc1 activation (48). NFATc1 is crucial to the differentiation of T_H cells to the T_H2 effector phenotype (see sect. VF1). Interestingly, disruption of SAPK or both SAPK and - leads to the preferential accumulation of T_H2 cells, consistent with the notion that SAPK-mediated inhibition of T_H2 differentiation is mediated through NFATc1 (48, 76, 77) (see sect. VF1).

C) MAX. Max is a 12-kDa helix-loop-helix (HLH) polypeptide that interacts with the transcription factor c-Myc enabling c-Myc to trans-activate at least a subset of its target genes. c-Myc is a key regulator of cell proliferation, differentiation, and apoptosis. The biological functions of c-Myc are thought to depend in part on the polypeptide binding partners with which c-Myc interacts and the regulation of these binding partners (20). A COOH-terminally truncated hnRNA splicing isoform of p38, referred to as Max-interacting protein-2 (Mxi2), was isolated in a yeast two-hybrid screen for Max interactors. Max and either Mxi2 or p38 can form a tight complex in vivo and in vitro, an interaction thought to be mediated by the HLH domain of Max and a similar loop on Mxi2 and p38. Max is a good substrate for both Mxi2 and p38 (363). Interestingly, there are no consensus proline-directed sites on Max, suggesting that the absolute proline requirement for MAPKs may be less pronounced for p38. The functional significance of p38-catalyzed Max phosphorylation is unclear (363) (Fig. 3).

D) ACTIVATOR PROTEIN-1. The SAPKs and p38s are the dominant stress-activated Ser/Thr kinases responsible for the recruitment of the activator protein-1 (AP-1) transcription factor (147, 165). ERK5 also plays a role in stress-activated AP-1 regulation (150). AP-1 is a heterodimer comprised of bZIP transcription factors, typically c-Jun and JunD, along with members of the fos (usually c-Fos) and ATF (usually ATF2) families. All bZIP transcription factors contain leucine zippers that enable homo- and heterodimerization, and AP-1 components are organized into Jun-Jun, Jun-Fos, or Jun-ATF dimers (147) (Fig. 4).

View larger version (50K): [in this window] [in a new window]   Fig. 4. Docking sites and stress-activated protein kinase (SAPK)/p38 substrate specificity. A: SAPKs bind to the c-Jun  domain, enabling phosphorylation of transcription factors with which c-Jun is dimerized, even if these dimerization partners do not possess  domain regions capable of binding SAPKs. B, top: proline-directed phosphoacceptor motifs of c-Jun family members. Note JunB has no proline-directed phosphoacceptor sites (underlined in c-Jun and JunD), and is therefore not a SAPK substrate. Bottom: hydrophobic SAPK docking sites in Jun family members. Note that despite the sequence similarities, only c-Jun  and JunB can associate with SAPKs. C: illustration of the selectivity of MAPKs for substrates mediated by the CD motif of MAPKs and the polybasic docking sites (DS in the figure) for substrates. The binding and phosphorylation of p38 by MKK6 and MNK1 by p38 are shown. D, top: polybasic docking sites of MKK6 and MNK1. Bottom: representative CD motifs of MAPKs (for details see Ref. 301).

The presence of Jun family members enables AP-1 to bind to cis-acting elements containing the 12-O-tetradecanoylphorbol 13-acetate (TPA) response element (TR; consensus sequence: TGA^C/_GTCA). ATFs, including ATF2, are members of the CREB subfamily of bZIP transcription factors. AP-1 heterodimers containing ATF transcription factors can bind both to the TRE and to the CRE (113, 147). AP-1 is an important trans-activator of a number of stress responsive genes including the genes for IL-1 and -2, CD40, CD30, TNF, and c-Jun itself. In addition, AP-1 participates in the transcriptional induction of proteases and cell adhesion proteins (e.g., E-selectin) important to inflammation (147, 252).

Activation of AP-1 involves both the direct phosphorylation/dephosphorylation of AP-1 components as well as the phosphorylation and activation of transcription factors that induce elevated expression of c-jun or c-fos. Both events can be activated independently by several signaling pathways (Fig. 4). Thus c-Jun is phosphorylated in resting cells at a region immediately upstream of the COOH-terminal, the DNA binding domain (Thr-231, Thr-239, Ser-243, Ser-249). This phosphorylation inhibits DNA binding and can be catalyzed in vivo by either glycogen synthase kinase-3 (GSK3) or casein kinase II (CKII). Upon stimulation of cells with AP-1 activators, the c-Jun COOH-terminal phosphates are removed under conditions wherein GSK3 is inactivated (24, 147). GSK3 is inhibited upon mitogen stimulation by a mechanism dependent on PI 3-kinase (58, 79).

Phosphorylation of c-Jun or ATF2 within their NH₂-terminal trans-activation domains correlates well with enhanced trans-activating activity (68, 112, 166, 244). The SAPKs can phosphorylate the c-Jun trans-activating domain at Ser-63 and Ser-73 (68, 166, 244). These residues are phosphorylated in vivo under conditions wherein the SAPKs are activated. Immunodepletion of SAPK from cell extracts removes all stress- and TNF-activated c-Jun kinase (166, 237). Thus SAPKs are the dominant kinases responsible for stress- and TNF-activated c-Jun phosphorylation (Fig. 2). JunD is also phosphorylated at Ser-90 and Ser-100 by SAPKs, albeit less effectively than is c-Jun. Ser-90 and Ser-100 of JunD lie within a region of the JunD trans-activation domain similar to the phosphoacceptor domain of c-Jun (144).

The SAPKs and p38s can both phosphorylate ATF2 at Thr-69 and Ser-71 in the trans-activation domain. Again, these residues are phosphorylated in vivo under conditions in which the SAPKs and p38s are activated. Phosphorylation of ATF2 at Thr-69 and Ser-71 correlates with activation of ATF2 trans-activating activity (112) (Fig. 3). Whether the SAPKs or p38s represent the dominant ATF2 kinases depends on the cell type and stimulus used. During reperfusion of ischemic kidney, for example, the SAPKs are the only detectable ATF2 kinases (219), whereas in KB keratinocytes treated with IL-1, the p38s are the major ATF2 kinases (60).

The SAPKs, p38s, and ERK5 (primarily in response to stress) and the ERKs (primarily in response to mitogens) also contribute to AP-1 activation by stimulating the transcription of genes encoding AP-1 components (147, 150) (Fig. 3). Induction of c-fos expression is one of the earliest transcriptional events known to occur. The fos promoter includes a cis-acting element, the serum response element (SRE), that mediates the recruitment of a heterodimeric transcription factor containing two polypeptides, the serum response factor (SRF) and a member of the ternary complex factor (TCF) family (318) (Fig. 3).

The TCFs include Elk-1 and Sap-1a (318). The SAPKs, and ERKs, but not p38, can phosphorylate two critical residues in the Elk1 COOH terminus (Ser-383, Ser-389), while the p38s can efficiently phosphorylate the corresponding residues (Ser-381 and Ser-387) on Sap1a (104, 138, 194, 337, 351). This phosphorylation enhances the binding of TCFs to the SRF and thereby triggers trans-activation at the SRE. By these processes, MAPKs activated by both stress and mitogens can convergently contribute to c-fos induction (Fig. 3).

The p38s and ERK5 can also phosphorylate the transcription factors myocyte enhancer factor-2 MEF2 subgroup of the MCM1-agamous and deficiens-SRF (MADS) box transcription factor family (116, 150, 365). MEF2s (MEF2A-D) were originally identified as transcription factors that bound to AT-rich sequences (consensus: CTAAAAATAA) and trans-activated key genes involved in myoblast differentiation; however, some MEF2s, notably MEF2C, are widely expressed and may regulate numerous other transcriptional regulatory events (94). Only MEF2A and -C are MAPK substrates. MEF2B and -D are not MAPK substrates; however, MEF2B and -D may act in conjunction with phosphorylated MEF2A/C. Thus phosphorylation enhances the trans-activating activity of MEF2A and -B or A and D heterodimers (365). MEF2A is phosphorylated at Thr-312 and Thr-319 by p38. Phosphorylation by other MAPKs has not been observed (365). MEF2C phosphorylation by MAPKs is apparently more complex. The p38s and ERK5 phosphorylate different sets of sites on the MEF2C polypeptide. Thus p38s phosphorylate Thr-293 and Thr-300, whereas ERK5 phosphorylates Ser-387 (116, 150). All three sites are phosphorylated in response to serum or stress; however, Thr-293/Thr-300 phosphorylation is sufficient for p38 activation of MEF2C while Ser-387 phosphorylation is sufficient for ERK5 activation of MEF2C (116, 150). A cis-element for MEF2C resides in the c-jun promoter; thus p38 and ERK5 activation can contribute to the induction of c-jun expression (116) which, in turn, potentiates to AP-1 activation. Indeed, MEF2A or -C, once activated by p38s, can trans-activate the c-Jun promoter (116, 150, 365) (Fig. 3).

The phosphorylation of c-Jun by the SAPKs highlights an important point about the mechanism of substrate recognition by members of the MAPK family. All MAPKs are "proline-directed," phosphorylating Ser/Thr residues only if followed immediately by proline (see sect. I and Fig. 4). However, the specificity of MAPKs for their physiological substrates is dictated in large part by the presence of binding sites, often substantially distal from the phosphorylation sites, that are specific for distinct MAPK subgroups. These MAPK binding sites allow for the selective interaction between MAPKs and their true in vivo substrates.

Two general classes of MAPK binding site have been described for MAPK substrates. The first of these to be identified was a hydrophobic cluster of residues present in many transcription factor substrates of MAPKs. Thus the SAPKs, but not the ERKs or p38s, bind c-Jun quite strongly. The SAPK binding site on c-Jun lies between residues 32 and 52, well away in the primary sequence from Ser-63 and Ser-73, the sites of phosphorylation (63, 143, 144, 147). The SAPK docking site overlaps with the so-called -domain (amino acids 30-57), a hydrophobic region initially implicated in the regulation of c-Jun oncogenicity due to its deletion in oncogenic v-Jun (accordingly, v-Jun does not bind SAPK and is not a SAPK substrate) (63, 143, 144, 147) (Fig. 4).

The binding of c-Jun to SAPK isoforms has been mapped to a -strand-like region spanning subdomains IX and X. Interestingly, a portion of this region undergoes differential hnRNA splicing in type I and II SAPKs (63, 111, 143, 166) (Table 1 and sect. IIE4). This -strand region also differs substantially among the three SAPK genes, perhaps accounting for differential SAPK substrate selectivity. Thus SAPK interacts most strongly with c-Jun (63, 111, 143).

The presence of the SAPK binding site, in conjunction with the ability of c-Jun to heterodimerize with other members of the Jun family, enables the SAPKs to phosphorylate other AP-1 constituents in vivo that, as monomers or homodimers, are poor SAPK substrates (144, 147). Thus JunD possesses domains similar to the phosphoacceptor and SAPK binding domains of c-Jun. In spite of this, JunD binds SAPK poorly (144) (Fig. 4). Accordingly, purified JunD is not ordinarily a SAPK substrate in vitro; however, when heterodimerized with c-Jun, JunD can undergo efficient SAPK-catalyzed phosphorylation in vitro, and activation in transfection experiments in vivo (144) (Fig. 4). JunB also possesses a conserved region homologous to the SAPK binding pocket of c-Jun, and in contrast to JunD, JunB binds SAPK well. However, JunB does not possess the proline-directed phosphoacceptor sites that are required for SAPK phosphorylation (144, 147) (Fig. 4A). Thus JunB is not a SAPK substrate in vivo or in vitro (144) (Fig. 4B). However, in transfection experiments, JunB can heterodimerize with c-Jun mutants missing the -domain and, because JunB can bind SAPK, foster SAPK phosphorylation of these mutants (144) (Fig. 4B).

ATF2 also contains a hydrophobic pocket (amino acids 20-60) similar to the c-Jun -domain that binds SAPKs and p38s. As with the c-Jun -domain, the ATF2 MAPK binding site lies NH₂ terminal to the phosphoacceptor sites (Thr-69 and Thr-71) (112). Likewise, Elk-1 has a MAPK docking site, the D-domain (amino acids 312-334) that lies NH₂ terminal to the phosphoacceptor sites (Ser-383, Ser-389) (337, 351). Interestingly, the Elk-1 D-domain is essential for ERK and SAPK binding and phosphorylation; however, p38 binding and phosphorylation, the physiological significance of which has not been unambiguously established (138, 351), appears not to require this domain (351). The role of the SAPK -strand in determining the differential binding of different SAPK isoforms to ATF2 or Elk1 is unknown.

A second class of MAPK binding domain generally consists of a small stretch of basic residues (Fig. 4C). This domain is prevalent among protein kinase substrates of MAPKs, including MAPKAP-K1s/Rsks (ERK substrates), MNKs, MAPKAP-K2/3, and PRAK and binds to the recently identified common docking (CD domain) domain, an extracatalytic region rich in acidic residues, found in all MAPKs (301). Interestingly, the basic MAPK binding motif is found not only in many MAPK substrates, but in a diverse array of MAPK regulators including MEKs (notably the SAPK-specific MEKs SAPK/ERK kinase-1, SEK1 and some MAPK-kinase-7, MKK7, isoforms, as well as the p38 MEKs MKK3 and MKK6) and MAPK phosphatases (301) (Fig. 4C). It is likely that the basic residues in these novel MAPK binding sites interact electrostatically with the cognate MAPK CD motifs. Moreover, differences among the basic MAPK docking sites and MAPK CD domains likely confer a high degree of specificity among MAPKs for their substrates and activators. This is readily apparent in the specific activation of MAPKAP-K2, under initial rate conditions, by p38 (see sect. IIE1), and in the highly selective scaffolding function of SEK1 (see sect. IIIC4).

As discussed in section IIB, the SAPKs are encoded by three genes that undergo differential hnRNA splicing to generate at least 10, and possibly up to 12, polypeptide species (111, 166). As noted in section IIE3, there is some evidence that the hnRNA splicing in the SAPK -strand region spanning kinase subdomains IX-X that generates the type 1 and type 2 enzymes (Table 1) may affect the affinity of different SAPK splicing isoforms for substrate. Thus SAPK-p542 (JNK21, Table 1) binds more strongly to c-Jun than does SAPK-p541 (JNK22, Table 1). Moreover, SAPK type 2 (JNK2 type , Table 1) isoforms bind more strongly to c-Jun than to ATF2, whereas SAPK type 1 (JNK2 type ) isoforms bind more strongly to ATF2 than to c-Jun (63, 111, 143).

In addition to differences in substrate binding among splicing isoforms of the same SAPK gene, the different sapk gene products themselves, SAPK, -, and - (JNKs-2, -3 and -1, respectively, Table 1), may also display differential substrate selectivity. Thus, whereas all SAPK isoforms bind more strongly to c-Jun compared with ATF2, SAPK/JNK2 isoforms are overall higher affinity c-Jun kinases than are either SAPK/JNK3 or SAPK/JNK1, a feature likely attributable to the high affinity for c-Jun conferred by the -strand region in domains IX-X of the SAPK -isoforms (Table 1) (63, 111, 143). It should be noted, however, that these differences as well as those among the splicing isoforms are modest (2- to 5-fold in vitro), and it is unclear what the in vivo biochemical significance of these differences is. These relatively subtle effects may have more profound ramafications at the cellular and organismal level. As we shall see in section V, more dramatic differences among SAPK isoforms are observed when the biological functions of these enzymes are examined.

The different aspects of AP-1 regulation, activation of constituent transcription factor expression and direct phosphorylation/activation of constituent transcription factors, can be independently regulated by several pathways in response to different types of stimuli (Fig. 9). Thus mitogenic stimuli, which preferentially recruit the ERKs and inhibit GSK3, will preferentially activate AP-1 through enhancement of expression of AP-1 components (via ERK phosphorylation of Elk-1, resulting in c-fos expression, for example), and through the relief of GSK3-mediated inhibition of c-Jun DNA binding (Fig. 3).

In contrast, stresses and inflammatory cytokines such as TNF, which preferentially activate the SAPKs and p38s, can recruit AP-1 through the direct phosphorylation of AP-1 components (c-Jun by the SAPKs and ATF2 by both the SAPKs and p38s). However, stress pathways can also promote enhanced expression of AP-1 components through recruitment of Elk-1 (mediated by SAPK phosphorylation), which results in elevated c-fos expression, and through p38- and ERK5-catalyzed phosphorylation of MEF2A/C (Fig. 3). MEF2A/C can bind and trans-activate the promoter for c-jun. Stresses can also modestly recruit PKB/Akt, which can act to inhibit GSK3, thereby blocking its negative regulation of c-Jun DNA binding. Finally, the c-Jun promoter also contains an AP-1 site; thus c-jun expression can be autoregulated by any pathway that activates AP-1 (Fig. 3).

The ERKs, SAPKs, and p38s were first purified from cytosolic extracts and, indeed, a substantial subset of MAPK substrates are cytosolic (95, 117, 164, 174, 251, 259). Still, a great many nuclear substrates for MAPKs have been identified, and the critical question is how MAPKs localized in the cytosol could regulate nuclear proteins such as transcription factors.

Of the mammalian MAPKs, the regulation of ERK nucleocytoplasmic shuttling has been studied the most extensively, and it is likely that many of the characteristics of the regulation of ERK subcellular localization may be applicable to stress-regulated MAPKs. Early immunocytochemical examination of the subcellular localization of ERKs revealed that upon mitogen stimulation, a substantial portion of the pool of ERK immunoreactivity translocates to the nucleus. Similar stimulus-dependent translocation of the SAPKs and p38s was subsequently observed (31, 40, 247). This translocation is reversible and terminates upon cessation of stimulus (40). The potential importance of ERK nucleocytoplasmic shuttling became evident when it was observed that ERK-dependent de novo gene expression correlated with prolonged ERK activation, conditions that coincided with nuclear translocation (26, 74, 317).

Nuclear translocation is essential for ERK-dependent activation of gene expression and regulation of the cell cycle. SAPK and p38 translocation may be similarly important. Thus sequestration of ERK in the cytosol, by coexpression with a catalytically inactive form of the MAPK phosphatase (MKP) MKP3, has no effect on ERK-dependent Rsk activation or phosphorylation of an engineered cytosolic mutant of Elk-1, but strongly inhibits ERK-dependent gene transcription and S phase entry (26).

The molecular mechanism by which ERK nuclear localization is regulated is still incompletely understood. Several novel findings suggest that inactive ERK is retained in the cytosol as part of a complex with its immediate upstream activator MEK1. Formation of this complex requires a polybasic amino acid ERK binding domain located within the NH₂-terminal 32 amino acids of MEK1 (97, 301). As noted above, similar binding site for SAPK exists on at least a subset of SAPK-specific MEKs including SAPK/ERK kinase-1 (SEK1), MAPK-kinase-7 (MKK7), as well as the p38-specific MEKs MKK3 and MKK6 (see sects. IIE3, IIF1, IIF2, and IIIC4). These MAPK binding sites correspond to the basic motifs present on many MAPK protein kinase substrates (301). MEK1 itself is retained in the cytosol by a consensus nuclear export signal (amino acids 33-44). In addition, MEK cytosolic localization may be maintained by scaffold proteins, several of which have been identified for the SAPKs (see sect. IIIC). Dissociation of the MEK-ERK complex requires MEK-catalyzed phosphorylation of ERK at the Thr-Glu-Tyr phosphoacceptor region. Once dissociation of the complex occurs, activated ERK translocates to the nucleus (97).

Whereas ERK nuclear translocation requires phosphorylation to enable dissociation from MEK1, ERK catalytic activity is not required for translocation. Both wild-type and kinase-inactive (Lys45Arg) ERK2, phosphorylated in bacteria upon coexpression with activated MEK1, could translocate to the nucleus upon microinjection into cells. In contrast, mutation of the ERK2 phosphoacceptor sites to nonphosphorylatable residues (Thr185Ala/Tyr187Phe) completely abrogated the ability of these kinases to translocate upon injection into fibroblasts (158).

Once the MEK-ERK complex dissociates, there is some evidence that a portion of the pool of ERK (primarily an active fraction that is not dimeric, see below) can freely diffuse into the nucleus. However, a substantial pool of ERK is translocated into the nucleus as part of a tightly regulated mechanism (3, 97, 158). ERK2 activation-dependent dimerization is also critical for translocation. Thus the crystal structure of activated ERK2 revealed a dimer with the interface occurring to the rear of the catalytic cleft between the two kinase lobes (28, 158). Dimerization requires four Leu residues (Leu-333, Leu-336, Leu-341, Leu-344) plus His-176 that forms a salt bridge with Glu-343 (28, 158). Mutation or deletion of the critical residues of the dimer interface has no effect on the in vitro kinase activity or MEK-catalyzed phosphorylation of the ERK2, but abolishes completely ERK2 nuclear translocation (158). Thus phosphorylation and activation of ERK2 results in dissociation from MEK1 and dimerization, both of which are necessary for translocation. It is noteworthy that both the SAPKs and p38s have similar dimerization motifs and show similar ligand-dependent translocation (31, 158, 247). Indeed, as is described in section IIIC4, SAPKs form reversible, activation-dependent associations with the SAPK-specific MEK SEK1, complexes that may serve in part to retain inactive SAPKs in the cytosol (301, 342). The ERKs, SAPKs, and p38s do not possess consensus nuclear localization motifs, and the mechanism by which ERKs traverse the nuclear pore complex is unclear. It is conceivable that MAPKs, once liberated from elements restraining them in the cytosol, move into the nucleus by interacting with nuclear substrate proteins.

F.  MEKs Upstream of the SAPKs, p38s, and ERK5

1.  Activation of the SAPKs by SEK1 and MKK7

The SAPKs are activated upon concomitant phosphorylation at Thr-183 and Tyr-185. The MEK SAPK/ERK kinase-1 (SEK1, also called MAPK-kinase-4, MKK4; MEK4; JNK kinase-1, JNKK1; and SAPK-kinase-1, SKK1, Table 2) was cloned independently by two groups who employed degenerate PCR to identify novel MAPK signaling components. Two potential initiation methionines, separated by 34 amino acids, are present in the SEK1 cDNA sequence, raising the possibility that two SEK1 isoforms exist that differ at their extreme NH₂ termini. Alternative splicing isoforms of SEK1 have not been identified in vivo; however, deletion of the NH₂-terminal 34 amino acids of SEK1 could affect its scaffolding function and its regulation by the MAP3K MEK kinase-1 (MEKK1) due to loss of a MEKK1 interaction motif and a basic amino acid-rich SAPK binding site (see sects. IIE3 and IIIC4) (69, 266, 301, 342).

The homology shared by SEK1 and MEKs-1 and -2 (as well as yeast MEKs) indicated that this kinase lay upstream of MAPKs. It was shown subsequently that SEK1 could phosphorylate and activate all three SAPK isoforms in vivo and in vitro (69, 266). Dérijard et al. (69) also showed that SEK1 could phosphorylate and activate p38 in vivo, when overexpressed, and in vitro (69). However, the significance of p38 activation by SEK1 is unclear. Targeted disruption of sek1 in mice has no effect on p38 activity in ES cells (226). Moreover, if SEK1 concentrations in in vitro assays are adjusted to initial rate conditions for SAPK activation, little or no p38 activation is observed (60, 205).

While the identification of SEK1 was encouraging, much biochemical evidence indicated that SEK1 was not the only SAPK-activating MEK. These studies revealed that the spectrum of SAPK activators recruited by different stimuli depended on the stimulus used and on the cell type. Thus fractionation on hydroxylapatite columns of extracts of osmotically shocked 3Y1 fibroblasts demonstrated a broad peak of SAPK activating activity that was fully resolved from a separate peak of SEK1 immunoreactivity (216). Likewise, Mono-S chromatography of KB cell extracts showed that IL-1 failed to activate SEK1, but stimulated a broadly eluting peak of SAPK activating activity distinct from SEK1. Similar multiple peaks of SAPK activating activity were observed in extracts of PC-12 cells treated with arsenite or osmotic shock and in KB cells treated with osmotic shock, UV radiation, or anisomycin (205). In osmotically shocked 3Y1 fibroblasts, SEK1 represented a comparatively minor peak of SAPK activating activity (216). In contrast, SEK1 was more strongly activated by osmotic shock, UV radiation, and arsenite in PC-12 cells (205). In KB cells, SEK1 and other SAPK activators were activated by anisomycin, osmotic shock, and UV radiation (205).

Genetic studies lent further support the contention that multiple SAPK-specific MEKs existed. Thus targeted disruption of sek1, while embryonically lethal, results in only a partial ablation of SAPK activation; sek1 / ES cells are refractile to anisomycin and heat shock activation of SAPK, while osmotic shock and UV activation of SAPK are unaffected (226). Finally, hemipterous is a Drosophila MEK required for dorsal closure during embryogenesis, and deletion/mutagenesis of hemipterous is lethal (see sect. IVC1). Although hemipterous is significantly homologous to SEK1, SEK1 cannot rescue Drosophila mutants wherein hemipterous is deficient (105, 124).

MKK7 (also called MEK7, JNKK2, and SKK4, Table 2) was isolated contemporaneously by several laboratories. Two of the strategies used, database mining or degenerate PCR, sought mammalian MEKs with close homology to hemipterous (91, 188, 218, 314, 340, 353). Alternatively, MKK7 was cloned in a two-hybrid screen as a polypeptide that could associate in vivo with MEK1 (124). The significance of this association is unclear. Consistent with its structural homology to hemipterous, MKK7 can effectively rescue hemipterous lethality, whereas SEK1 cannot (124). MKK7 displays a strong preference for SAPK, even under conditions of high expression. This contrasts with SEK1 which can, under conditions of high overexpression, activate p38 (69, 91, 188, 218, 314, 340, 353). MKK7 can activate all SAPK isoforms tested equally well (91, 188, 218, 314, 340, 353). MKK7, like SEK1 is subject to alternative hnRNA splicing yielding enzymes with three different NH₂ termini (, , and ) and two different COOH termini (types 1 and 2). The - and -isoforms bind directly to SAPKs via an NH₂-terminal extension not present in the -isoforms. Upon overexpression, the -isoforms exhibit a lower basal activity and higher activation by upstream stimuli (315).

MKK7 is strongly activated by TNF and IL-1, conditions which, at best, induce modest SEK1 activation. In contrast, SEK1 and MKK7 are activated with equal potency by osmotic shock while MKK7 is more weakly activated by anisomycin than is SEK1 (91, 188, 218, 314, 340, 353). Thus it is plausible to argue that MKK7 represents at least a substantial portion of the biochemically detected SAPK activating activity present in 3Y1, PC-12, or KB cells subjected to osmotic shock or anisomycin, or in KB cells treated with IL-1.

Although the activation patterns of SEK1 and MKK7 appear not to overlap entirely, there is evidence that SEK1 and MKK7 actively cooper